The gain from three temperature indices was uncertain in 26 Adventdalen reindeer count estimates for 2000–2025

We tested whether seasonal temperature improved later summer count estimates for reindeer at Adventdalen, Svalbard. We used NPI MOSJ counts from 1979–2025 and NOAA daily temperature at Svalbard Airport. The inputs describe the previous summer, October, and winter warmth through March. The baseline uses past count and a time trend. A second model adds the three temperature inputs. Each target from 2000–2025 was estimated with models fitted to earlier years only. Added temperature reduced mean squared log error by 4.56% across 26 targets. The 95% three-year block resample range was −39.87% to +28.90%. Mean absolute log error rose from 0.16599 to 0.16666. A check that refitted the models during resampling gave a wider gain range of −138.67% to +35.46%. Without a time trend, the apparent temperature gain was 32.06%, with a range that still included zero. Thus, these inputs gave no clear added gain beyond past count and time in this test. The winter input does not measure icing. The result does not reject known climate effects on reindeer. It applies to one count series, one station, and the stated models. The package has source data, code, all results, figures, hashes, and independent checks.

Svalbard · reindeer · population monitoring · temperature · temporal validation · ecological prediction · open data